📊 Data Science, Statistics & Business Intelligence
Every school, shop and farm produces numbers. This department teaches how to collect them cleanly, understand them honestly, and present them clearly — so decisions rest on evidence rather than guesswork.
A team reviewing its figures together. Photo: Startup Stock Photos · CC0 · source
Working with data on a laptop. Photo: Startup Stock Photos · CC0 · source
Five guided solutions
Each solution follows the Global Solutions pattern: the problem, how it works, the steps, the outcome you can expect — and the boundary where a licensed professional must take over.
Solution 1 of 5
Start with clean data
The problem. Every analysis fails the same way: messy inputs. Names spelled five ways, missing dates and mixed units quietly poison every chart built on them.
How it works. Cleaning is unglamorous, decisive work: one format for dates, one spelling for categories, blanks investigated rather than ignored. An hour of cleaning saves days of arguing about wrong conclusions.
Practical steps
List your columns and decide the one correct format for each.
Standardise category spellings — 'Kumasi' and 'kumasi ' are two cities to a computer.
Hunt duplicates and decide, by rule, which record survives.
Mark missing values visibly instead of silently typing zeros.
Save the cleaned copy separately; never edit your only original.
Expected outcome. Data you can trust — the foundation every honest chart stands on.
⚠ Professional boundary. Personal data carries legal and ethical duties: collect the minimum, store it securely, and never share what identifies people without consent.
Solution 2 of 5
Averages that don't mislead
The problem. 'Average' hides as much as it reveals: one very large value can drag a mean far from what is typical, making a school, shop or ward look richer or poorer than it is.
How it works. Use the trio together: the mean (total shared equally), the median (the middle value) and the range (smallest to largest). When mean and median disagree, the gap itself is information — a few extremes are pulling the story.
Practical steps
Compute the mean: total divided by count.
Sort the values and find the median — the middle one.
Note the smallest and largest values beside every average you report.
Ask who or what the extremes are before publishing any conclusion.
Choose the measure that matches the question: typical experience → median; totals and budgets → mean.
Expected outcome. Numbers that describe reality instead of decorating a guess.
⚠ Professional boundary. Never present an average without its count — an average of three students and of three thousand deserve different confidence.
Solution 3 of 5
Build a one-page dashboard
The problem. Decision-makers drown in tables and starve for the five numbers that matter, visible at a glance and always current.
How it works. A good dashboard is an act of editing: choose the few indicators that trigger action, show each against its target or its past, and let everything else live one click deeper.
Practical steps
Write the decisions this dashboard must support — if none, stop.
Choose at most six indicators, each with a clear owner.
Show each indicator with its comparison: target, last period, or both.
Use simple charts — bars for categories, lines for time — and label everything in plain words.
Fix a refresh rhythm (daily, weekly, monthly) and honour it.
Expected outcome. A single page that starts the right conversations every week.
⚠ Professional boundary. A dashboard shows what is happening, not why. Pair it with the habit of asking why before acting.
Solution 4 of 5
Read trends and seasons honestly
The problem. Sales rise every December and school fees bite every term — mistaking a season for a trend leads to hiring, borrowing and stocking at exactly the wrong moments.
How it works. Compare like with like: this month against the same month last year, not against last month. Draw at least a full year before claiming a direction, and mark known events (a promotion, a closure, a storm) on the chart where they happened.
Practical steps
Gather at least twelve periods of data before judging direction.
Compare each period with the same period a year earlier.
Annotate the chart with events that explain spikes and dips.
Separate the seasonal pattern from the underlying movement.
State your uncertainty: 'rising for three months' is a fact; 'will keep rising' is a hope.
Expected outcome. Plans built on the real rhythm of the work, not on one exciting month.
⚠ Professional boundary. Forecasts are estimates with error attached. Never borrow or commit against a forecast as if it were a fact.
Solution 5 of 5
Ask the right question first
The problem. Analysts waste weeks answering the question that was asked instead of the one that matters: 'sales fell' is a symptom; the real question is where, for whom, and since when.
How it works. Frame every analysis as a decision: What will we do differently depending on the answer? Who is affected? Over what period? A question with a decision attached produces analysis that gets used.
Practical steps
Write the decision the analysis must inform, in one sentence.
Define the population and period precisely.
Break the big question into measurable sub-questions.
Agree what evidence would change the decision — before seeing the data.
Report the answer, the evidence and the limits together.
Expected outcome. Analysis that changes actions, not analysis that decorates reports.
⚠ Professional boundary. If the honest answer is 'the data cannot tell us', say so — inventing certainty is the one unforgivable analytic sin.
